Last June, when I decided to pick up blogging again after four years of COVID life with two babies at home, I knew the only way to rebuild my writing habit was to commit to a regular publishing routine.

As I'd learned years earlier, a near-daily habit of writing and publishing is an incredibly effective way to get over the fear of hitting "publish." Even if you knock out a banger of a blog post one day, you still show up with something else to say the next day. Over time—and with a little help from AI—100+ blog posts later, I can see how much my writing has improved.

Creating things also helps solidify memory and meaning. Over the past six months, I’ve often found myself calling back to past posts, concepts or themes almost daily. Writing has become such a habit that even when I try to stop, I can’t.

But I've been working on building another habit too: AI fluency. Learning AI is less like picking up a tool and more like learning a language—you improve through practice. And that means making stuff.

Now that I've graduated out of "custom GPT" builder mode into basic web app builder mode, I'm continuing to improve this muscle of practicing the build. This week, I wanted to learn how to deploy something on Vercel, and I wanted to practice remixing my own content in an automated way.

Here's what happened.

Earlier this week, the Paragraph.xyz team (where this blog is published) previewed a new landing page look for many of us regular writers on the platform. I was excited to see that part of this update included access to your blog's RSS feed.

I've been daydreaming for months about what type of AI-enabled "content remixes" I could offer as alternative content streams to my prolific source material. So, inspired by yesterday's nostalgic 1990s mixtape vibes and my newfound AI no-code skills, I asked myself: "What would 13-year-old Bethany do with an RSS feed of her blog, if she could code?"

I decided she'd want to create an AI-powered haiku generator to turn any blog post into a short-burst haiku format. So yesterday, I made a prototype. You can check it out here. (warning: this is very drafty...and maybe do not look at this on mobile...)

How it works is–you just pick a blog post, and then it locates that post via the RSS feed, and then feeds the text through an AI (in this case, I used Claude's API, the Anthropic product), which spits out a haiku in return.

For example, here's a haiku generated from Monday's post about bounce houses.

You might think it's weird to show off a pretty mediocre example of a micro-app that looks a lot like what a 7th grader in the 1990s could code up. But when it comes to learning how to work with AI, it's that the art of improvement comes in the iterations. Pushing "publish" is the beginning, not the end. You end up learning a lot of little hacks along the way (for instance, yesterday I learned when Comic Sans font isn't available on your phone, it defaults to cursive).

That's why, through our 10-week AI Learning Accelerator with Decoded Futures this fall, we encouraged all of our participants to push publish often (and as early as possible). This led to some astounding results of impact in just a few short months.

The way that we encouraged this active learning was to structure each session as a combination of theory and practice. Here's what that looked like:

By the way, it turns out that another great adult learning hack is to teach something back to others while still learning it yourself. This might seem like an aggressive approach to learning—it forces you to grasp the material quickly enough to explain it to people just a few steps behind you. But it's an incredibly effective (and exhilarating) way to accelerate your learning curve.

And since I don't have another class lined up to teach (at least not yet), this blog will have to do. So I encourage you to think about what habit or fluency you're looking to bolster in yourself this year and play along, too. But, you know, if this blog is too much content for you, you're more than welcome to just stick with "haiku mode" for now. <3